PPPavol Partila

Pavol Partila

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Pavol Partila has a total of 69 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Non-invasive detection of obstructive sleep apnea from speech in a limited clinical dataset

    by Martina Skapova, Pavol Partila, Jaromir Tovarek, Tereza Lubojacka, Jana Slonkova, Ondrej Hruby, Samuel Genzor, Jan Mizera, Petr Mooz

    This study investigates non-invasive detection of obstructive sleep apnea (OSA) from daytime speech in a limited polysomnography-labeled clinical cohort. The objective was to evaluate whether speaker-verification-inspired representations can support robust speech-based OSA screening under constrained clinical data conditions. We compared conventional mel-frequency cepstral coefficient (MFCC)-based classifiers with i-vector systems using cosine-similarity, support vector machine (SVM), and neural-network (NN) back ends, and included an exploratory ECAPA-TDNN embedding experiment using a pretrained frozen speaker model with prototype-based cosine scoring. Speech recordings from 93 participants were collected under controlled conditions and evaluated using subject-independent validation. Supervised i-vector back ends achieved the strongest performance. The i-vector SVM system obtained the highest numerical accuracy of 96.3% and the strongest threshold-independent performance, with AUC-ROC of 0.987 and AUC-PR of 0.992, while the i-vector NN system showed closely comparable performance with a non-significant AUC-ROC difference from i-vector SVM. ECAPA-TDNN showed moderate class separability but did not outperform the i-vector systems. Demographic-only and speech-demographic fusion analyses indicated that sex, age, and body mass index carried predictive information but did not fully explain the best speech-based performance. These findings support speech-based OSA screening as a low-burden complementary approach, while emphasizing the need for larger, balanced, and externally validated cohorts.


  • Open Access

    Statistics of the Sum of Double Random Variables and Their Applications in Performance Analysis and Optimization of Simultaneously Transmitting and Reflecting Reconfigurable Intelligent Surface-Assisted Non-Orthogonal Multi-Access Systems

    by Bui Vu Minh, Phuong T. Tran, Thu-Ha Thi Pham, Anh-Tu Le, Si-Phu Le, Pavol Partila

    For the future of sixth-generation (6G) wireless communication, simultaneously transmitting and reflecting reconfigurable intelligent surface (STAR-RIS) technology is emerging as a promising solution to achieve lower power transmission and flawless coverage. To facilitate the performance analysis of RIS-assisted networks, the statistics of the sum of double random variables, i.e., the sum of the products of two random variables of the same distribution type, become vitally necessary. This paper applies the statistics of the sum of double random variables in the performance analysis of an integrated power beacon (PB) energy-harvesting (EH)-based NOMA-assisted STAR-RIS network to improve its outage probability (OP), ergodic rate, and average symbol error rate. Furthermore, the impact of imperfect successive interference cancellation (ipSIC) on system performance is also analyzed. The analysis provides the closed-form expressions of the OP and ergodic rate derived for both imperfect and perfect SIC (pSIC) cases. All analyses are supported by extensive simulation results, which help recommend optimized system parameters, including the time-switching factor, the number of reflecting elements, and the power allocation coefficients, to minimize the OP. Finally, the results demonstrate the superiority of the proposed framework compared to conventional NOMA and OMA systems.


  • Open Access

    People Detection Using Artificial Intelligence with Panchromatic Satellite Images

    by Peter Golej, Pavel Kukuliač, Jiří Horák, Lucie Orlíková, Pavol Partila

    The detection of people in urban environments from satellite imagery can be employed in a variety of applications, such as urban planning, business management, crisis management, military operations, and security. A WorldView-3 satellite image of Prague was processed. Several variants of feature-extracting networks, referred to as backbone networks, were tested alongside the Faster R–CNN model. This model combines region proposal networks with object detection, offering a balance between speed and accuracy that is well suited for dense and varied urban environments. Data augmentation was used to increase the robustness of the models, which contributed to the improvement of classification results. Achieving a high level of accuracy is an ongoing challenge due to the low spatial resolution of available imagery. An F1 score of 54% was achieved using data augmentation, a 15 cm buffer, and a maximum distance limit of 60 cm.


  • Open Access

    Statistical Evaluation of the Impacts Detection Methodology (IDM) to Detect Critical Damage Occurrences during Quay Cranes Handling Operations

    by Sergej Jakovlev, Tomas Eglynas, Valdas Jankunas, Miroslav Voznak, Mindaugas Jusis, Pavol Partila, Jaromir Tovarek

    During various marine container handling operations, performed mainly in larger-scale container terminals, containers get damaged regularly. Our previous studies showed that each physical impact results in some form of physical deformation of the backbone structure. Even at low accelerations, the spreaders of the quay cranes impact the containers with enough force to substantially bend the metal parts of the corners of the containers, when additional hooking procedures are required. This means that the first time resulted in the metal rods hitting the metal frame with an average 15-ton mass at the average speed of 1.7 m/s. The metal rods of the hooking mechanisms’ impact areas of the containers are structurally important, and each impact surely damages the containers, diminishing their total operational time. We have already proposed the Impacts Detection Methodology (IDM) and its application system, tested in Klaipeda City port, and it proved to be efficient in real-time operations, detecting concurrent impacts with each new handling cycle. In this paper, we provide a summarisation of a larger number of detections using the IDM, and as a result of this analysis, we have detected that more impact events happen when containers are taken from the upper parts of the ship, in comparison to the ones taken from the shafts. Results suggest that more critical events occur due to operator actions and experiences working with the machinery, yet the same operators tend to make fewer impact mistakes taking the containers from the shafts as the vertical cell guides tend to direct the movements and lower the levels of the natural sway of the spreader inside closed environments. This surely damages the metal infrastructure of the shafts, as seen in our previous study, but minimizes the chances of secondary impacts occurring during hooking.


  • Open Access

    Detecting Physical Impacts to the Corners of Shipping Containers during Handling Operations Performed by Quay Cranes

    by Sergej Jakovlev, Tomas Eglynas, Mindaugas Jusis, Miroslav Voznak, Pavol Partila, Jaromir Tovarek

    This study aims to address the problem of proper shipping container damage detection during the hooking procedure performed by quay cranes and their hooking mechanisms. We adopted the Impacts Detection Methodology (IDM), developed previously, to detect repeated impacts on the same areas of the container. These concurrent impacts indicate false hooking procedures, which result in dangerous metal deformations in these areas over short periods of operational time. This application intends to verify if this methodology is adaptable in real-life applications to detect these specific events. Our main results indicate that more than half of handling procedures are carried out with a higher risk of structural damage to the containers due to these repeated impacts, which can reach up to five concurrent impacts in some case studies.


  • Open Access

    Detection of Physical Impacts of Shipping Containers during Handling Operations Using the Impact Detection Methodology

    by Sergej Jakovlev, Tomas Eglynas, Miroslav Voznak, Mindaugas Jusis, Pavol Partila, Jaromir Tovarek, Valdas Jankunas

    The transportation of cargo inside shipping containers is a risky operation that requires constant monitoring activities and real-time operational actions. Yet, the detection of the real dynamics of the container and the surrounding infrastructure and extraction of true subsequent critical events is still an unresolved issue among engineers. In this paper, we analyze the new physical impact detection method, namely the Impact Detection Methodology (IDM), to detect the most obvious and force-dependent impacts from acceleration data, using the IoT sensor in an experimental environment using the heavy machinery of a seaport. By variating the threshold level, we have observed the changes in the number of impacts detected within three separate case studies. Results suggest that the optimal parameters tend to provide an adequate number of events, yet even the slightest change in the threshold level can increase or decrease the number of detected impacts in a non-linear fashion, making the detection harder, due to unforeseen external impacts on the dataset, the filtering of which is still the main priority of our future research.


  • Detection of speaker liveness with CNN isolated word ASR for verification systems

    by Martina Slivova, Miroslav Voznak, Jaromir Tovarek, Pavol Partila


  • Isolated Word Automatic Speech Recognition System

    by Martina Slívová, Pavol Partila, Jaromír Továrek, Miroslav Vozňák


  • Deep Learning Serves Voice Cloning: How Vulnerable Are Automatic Speaker Verification Systems to Spoofing Trials?

    by Pavol Partila, Jaromir Tovarek, Gokhan Hakki Ilk, Jan Rozhon, Miroslav Voznak

    This article verifies the reliability of automatic speaker verification (ASV) systems on new synthesis methods based on deep neural networks. ASV systems are widely used and applied regarding secure and effective biometric authentication. On the other hand, the rapid deployment of ASV systems contributes to the increased attention of attackers with newer and more sophisticated spoofing methods. Until recently, speech synthesis of the reference speaker did not seriously compromise the latest ASV systems. This situation is changing with the deployment of deep neural networks into the synthesis process. Projects including WaveNet, Deep Voice, Voice Loop, and many others generate very natural and high-quality speech that may clone voice identity. We are slowly approaching an era where we will not be able to recognize a genuine voice from a synthesized one. Therefore, it is necessary to define the robustness of current ASV systems to new methods of voice cloning. In this article, well-known SVM and GMM as well as new CNN-based ASVs are applied and subjected to synthesized speech from Tacotron 2 with the WaveNet TTS system. The results of this work confirm our concerns regarding the reliability of ASV systems against synthesized speech.


  • Research of Quay Crane Control Algorithm with Embedded Sway Control Sub-routine

    by Tomas Eglynas, Mindaugas Jusis, Sergej Jakovlev, Audrius Senulis, Pavol Partila, Saulius Gudas

    In this paper, an integrated autonomous quay crane control algorithm with embedded container swinging control sub-routine was developed with the defined parameter set. We have used PID controller with auxiliary PI with feedback and an S-shaped input signal. Experimental results show a 66% increase in container sway damping ratio when using the S-shaped input signal.